How to Start a Ghost Kitchen: The Complete 2026 Blueprint.

A ghost kitchen has no dining room, no host, and no walk-in traffic. Every dollar arrives through a screen: DoorDash, Uber Eats, Grubhub, your own Toast or Olo ordering page, or a third-party catering marketplace like ezCater. That means your customer is not "people who live nearby" in the restaurant sense. It is the narrower set of households and offices inside a 3 to 7 mile delivery radius who order delivery at least twice a month and who are scrolling a marketplace app at 6:45 p.m. on a Tuesday. In most metros that is a real but finite pool, and you are competing for it against every brick-and-mortar restaurant in the same radius that also lists on the same apps.

The honest revenue picture: a single-brand ghost kitchen in a medium-volume U.S. market is typically modeled at $30,000 to $150,000+ per month gross, or roughly $360,000 to $1.8 million+ annually. Treat that as a scenario range, not a census figure. No primary national dataset tracks ghost kitchen revenue separately, because these operations get folded into broader limited-service restaurant statistics. What actually drives where you land inside that range is order count per day, average ticket, the share of orders coming through commission-heavy marketplaces versus your own channel, and whether you run one brand or three off the same line.

The buyers who matter most fall into three groups with very different economics. Marketplace consumers deliver volume but cost you 15% to 30% in commission per order. Direct online orders through your own site deliver 3% to 5% payment processing and nothing else, but require you to generate the demand yourself. Office and small-event catering, booked through ezCater or direct, delivers $200 to $1,500 tickets with one drop-off and is the single most overlooked revenue line for delivery-only kitchens. A launch plan that assumes only the first group is a launch plan that runs on someone else's algorithm.

Two structural facts shape 2026. First, commissary operators such as CloudKitchens, Kitchen United successors, and dozens of regional shared-kitchen landlords have made the entry price genuinely low, which means your competitive set includes hobbyists and undercapitalized operators who will discount aggressively and then close. Second, marketplace consumers are far more brand-indifferent than restaurant diners. They filter by cuisine, rating, delivery time, and promotion. If your rating slips below roughly 4.5 stars, your placement and your order volume drop together, and no amount of food quality argument reverses it quickly.

Because no customer visits, siting stops being about visibility and becomes pure delivery geometry. What matters is driver density and drive time. You want a location where DoorDash and Uber Eats drivers already circulate, which means near existing restaurant clusters, not in an isolated industrial park where a courier waits 12 minutes for a $4 pickup fee and declines your orders. Model your realistic radius at 3 to 7 miles; beyond that, food quality degrades and platforms deprioritize you in customer feeds. Count households and daytime office population inside that radius before you count square feet.

The three viable structures carry very different risk. A shared commissary slot runs roughly $1,500 to $15,000 to enter and often month-to-month, with hood, grease interceptor, walk-in, and dish already permitted, so buildout is near $0 and you can open in 8 to 20 weeks. A private shell buildout costs $80,000 to $150,000+ and takes 4 to 9 months, because Type I hood, Ansul suppression, grease interceptor, floor drains, and an electrical service upgrade all require plumbing permits, fire marshal signoff, and a fresh Certificate of Occupancy. The third path, subleasing off-hours capacity from an existing permitted restaurant, is the cheapest entry of all but the least controllable: their equipment, their health permit, their schedule.

The decision that costs the most money is signing a lease before calling the county environmental health department. A shell that looks perfect can be zoned in a way that requires a conditional use permit at $0 to $2,000+ and 8+ weeks of hearings, or can lack the gas capacity for the fryer bank your menu assumes. Ask three questions before signing: is this address already permitted as a retail food facility, does an approved Type I hood exist with current cleaning records, and is there a grease interceptor sized for my drainage. Any "no" adds weeks and five figures.

One siting nuance specific to multi-brand operators: if you plan to run three virtual brands off one line, verify with the health department and each marketplace how they treat multiple trade names at a single permitted address. Some jurisdictions require each brand name to appear on the food establishment permit or on labeling; some platforms require a distinct storefront verification per brand. Sorting this out at siting time is cheap. Sorting it out after a platform suspends two of your three storefronts is not.

Start with the ticket, not the month. Take a $32 marketplace order. Food cost at 25% to 32% takes $8.00 to $10.24. Packaging, the sealed container, bag, sleeve, utensils, and label, takes $0.80 to $2.00 per order. That leaves a gross margin in the 60% to 75% band the industry models. Then the marketplace takes its commission at roughly 15% to 30%, or $4.80 to $9.60 on that same ticket. You are now at $10 to $18 of contribution before a single person is paid, before rent, and before any promotion you ran to win the order.

Labor is the next bite and it does not scale down gracefully. Even a lean shift needs 2 to 5 people at $14 to $30 per hour: a kitchen manager or shift lead at $18 to $30, cooks at $15 to $22, prep at $14 to $19, a packer or expediter at $14 to $20. A four-hour dinner rush staffed with three people at an $18 blended rate is $216 of labor. If that rush produces 40 orders, labor is $5.40 per order. If it produces 18 orders, labor is $12 per order and the order is underwater. This is why order density per labor hour, not revenue, is the metric that decides whether you survive.

Rent and platform-adjacent costs finish the calculation. Commissary rent typically shows up as a monthly slot fee, and standalone rent plus utilities on a hood-ventilated kitchen runs materially higher per square foot than dry retail because of gas, exhaust, and refrigeration load. Add POS and middleware subscriptions of roughly $1,500 to $6,000 to set up plus ongoing monthly fees, payment processing at 3% to 5% on direct orders, and refunds and chargebacks, which on delivery are absorbed by the operator far more often than most first-year owners expect. Net margin for successful operators is modeled at 5% to 15%; many first-year kitchens sit below that while volume stabilizes.

The lever that changes this arithmetic is channel mix. The same $32 order placed on your own Toast or Olo page costs you 3% to 5% in processing instead of 15% to 30% in commission, which is $4 to $8 more contribution on one ticket. Move 25% of volume from marketplace to direct and you have added roughly a point or two of net margin without cooking a single extra dish. Catering through ezCater or direct corporate accounts does the same thing at larger tickets: $200 to $1,500 per drop with one delivery instead of twenty.

A single-brand ghost kitchen opens with 2 to 5 people per shift. The non-negotiable roles are a kitchen manager or shift lead at $18 to $30 per hour who holds the ANSI-accredited Food Protection Manager certificate, one to three cooks or line cooks at $15 to $22, and a packer or order coordinator at $14 to $20 whose only job is checking bags against tickets and sealing them. Prep cooks at $14 to $19 come next as volume grows, and dishwashing or utility at $14 to $18 is often folded into prep at launch. A multi-brand operation running three storefronts off one line realistically needs 6 to 12+ people across the week.

The packer role is the one first-time owners cut, and cutting it is the most expensive staffing decision in this business. In a restaurant, a server catches a missing side. In a ghost kitchen, a missing side becomes a one-star review and a refund you absorb. One person dedicated to bag accuracy, labeling, and handoff during a four-hour rush costs roughly $56 to $80 and protects the rating that drives your entire order flow. Treat it as a margin position, not overhead.

For national wage context, BLS reports combined food preparation and serving workers, including fast food, at about $14.80 median and $15.27 mean hourly; cooks, fast food, at about $14.81 median and $15.54 mean; and cooks, restaurant, at about $17.87 median and $18.37 mean. Those are national baselines. Actual 2026 pay runs higher in large metros and in states with higher minimum wages, and wage levels above the BLS baseline are the most uncertain number in your model because local labor conditions dominate. Check three comparable local postings before you set a rate.

Bookkeeping, marketing, and platform management do not need a full-time hire at launch. Budget for part-time or outsourced help at $20 to $45 per hour equivalent, and pick one weekly two-hour block where someone reconciles marketplace payouts against your POS. Delivery platforms deduct commissions, promotions, refunds, and adjustments before depositing, and unreconciled payouts are where first-year operators discover, three months late, that their real commission rate was closer to 30% than the 15% on the contract.

Days 1 through 14: launch on one marketplace, not three. Pick whichever platform has the most order volume in your specific zip codes, accept its new-store placement boost, and use it to accumulate your first 50 to 100 reviews. Set your prices 15% to 20% above your direct-channel prices to absorb commission, and turn on the platform's standard new-customer offer. Do not buy promoted placement yet. Placement money spent before you hold a 4.6+ rating buys traffic that turns into bad reviews at scale, and those reviews follow you for months.

Days 15 through 45: stack the remaining marketplaces and open your own direct channel. Your own ordering page through Toast, Square, or Olo costs 3% to 5% in processing versus 15% to 30% in commission, so every order you move over is worth $4 to $8 more contribution on a $32 ticket. Drive it with an insert in every marketplace bag offering a discount code good only on your own site, plus a Google Business Profile set to delivery-only with your real hours and menu photos. Claiming that profile is free and it is how customers who search your cuisine plus "delivery near me" find you outside the apps.

Days 30 through 90: go after the two channels that do not depend on algorithms. First, office and small-event catering: list on ezCater, then hand-deliver sample trays with a printed price sheet to 25 to 40 offices, clinics, dealerships, and law firms inside your delivery radius. A $200 to $1,500 catering ticket with one drop-off carries a fraction of the delivery cost of the same revenue in individual orders. Second, repeat orders: text or email your direct customers a Tuesday or Wednesday offer, because those are the softest delivery days in most markets and incremental orders on a shift you are already staffing are nearly pure contribution.

Track exactly four numbers weekly for 90 days: orders per day by channel, average ticket, blended commission rate as actually deposited, and star rating on each platform. If orders per day flattens for three consecutive weeks while rating holds, your problem is demand and you need more channels. If rating slides while orders hold, your problem is packaging and hold time, and you fix it before spending another dollar on marketing. Most kitchens that close spent the marketing budget on the wrong one of those two problems.